The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Git commands and project-defined validation scripts to manage the repository and verify changes.
Evidence includes Git operations in SKILL.md: git status, git diff, git log, git add, git commit, and git push (subject to explicit user request).
Step 5 in SKILL.md involves discovering and running project-specific scripts, configuration, and CI checks (e.g., npm test, linters).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from the repository, such as file diffs and commit history, which creates a potential surface for indirect prompt injection.
Ingestion points: SKILL.md (Step 1) reads output from git diff and git log; SKILL.md (Step 4) and references/changelog.md read the local CHANGELOG.md file.
Boundary markers: No explicit delimiters are used for the input data, but the skill defines strict formatting rules for its output.
Capability inventory: SKILL.md (Step 6 and Step 7) allows for file staging, committing, and pushing changes. Step 5 allows execution of project-defined validation commands.
Sanitization: The skill includes safeguards in Step 1 to flag sensitive files (secrets, env files) and untracked files before staging, and requires explicit paths for git add to prevent accidental staging of unrelated content.

prepare-commit